html5-img
1 / 30

Earth Science Data Records (ESDR) of Global Forest Cover Change

Earth Science Data Records (ESDR) of Global Forest Cover Change. Chengquan Huang 1 John R. Townshend (PI) 1 , Jeffrey G. Masek 2 , Matthew C. Hansen 3 1 Department of Geography, University of Maryland, College Park, MD 20742

ashby
Download Presentation

Earth Science Data Records (ESDR) of Global Forest Cover Change

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Earth Science Data Records (ESDR) of Global Forest Cover Change Chengquan Huang1 John R. Townshend (PI) 1, Jeffrey G. Masek2, Matthew C. Hansen3 1Department of Geography, University of Maryland, College Park, MD 20742 2Code 614.4 - Biospheric Sciences, NASA Goddard Space Flight Center, Greenbelt, MD 20771 3Geographic Information Science Center of Excellence, South Dakota State University, Brookings, SD 57007

  2. Project Background • Funded primarily through NASA’s MEASURES program • Significant LCLUC funding • Multi-institute collaboration • University of Maryland • John Townshend (PI) • Chengquan Huang, SaurabhChannan, Joe Sexton, RaghuramNarasimhan, Min Feng, Kuan Song, Xiaopeng Song, DoHyung Kim, Danxia Song, Paul Davis, Sam Goward, et al. • NASA/GSFC and contractors • Jeff Masek, Robert Wolfe, Eric Vermote, • FengGao, Bin Tan, Greg Ederer, Jim Tucker, et al. • South Dakota State University • Matt Hansen, Peter Potapov

  3. CO2 CO2 Disturbance Regrowth Primary Goals • Develop global FCC products to support • Modeling • Climate • Carbon: REDD • Hydrology • Biodiversity and conservation • Demonstrate routine global FCC monitoring capability

  4. “Global” Land Survey (GLS) 1975 1990 2000 2005 1975 1990 2000 2005 Deliverables Forest cover change (MODIS) ESDR Landsat-MODIS consistency MODIS VCF (2000-2005) Fragmentation & change ESDR Forest cover change (Landsat) ESDR Surface reflectance ESDR

  5. Subsets and Aggregation Global Products Landsat/MODIS resolution Aggregation Complete subset for all WDPA protected areas 250 m – 5 km (MODIS-AVHRR) 5 km – 1 degree (model grids)

  6. GFCC-ESDR Processing Flow

  7. Key Requirements of the Approach • Mass processing capability • ~9000 images x 4 epochs • Radiometric adjustment/atmospheric correction • Geometry mostly satisfactory • Automated change mapping • Consistency and quality checking

  8. Linux clusters and servers Data management Job scheduler Processing Archive and distribution Mass storage Mass Processing Capability • LEDAPS (Landsat Ecosystem Disturbance Adaptive Processing System) • An adaption of MODAPS by NASA/GSFC (Masek) • Have been used to process a few thousands Landsatimages • UMD enhancement • ~9,000 Landsat images x 4 epochs (10 times more than can be handled by current LEDAPS)

  9. Atmospheric Adjustment • Based on the MODIS 6S approach (Masek et al. 2006) • Fully automated • Validated at many locations After adjustment Before adjustment

  10. Landsat SR Mosaic for North America

  11. Legend 2 R Landsat-MODIS SR Comparison • Every Landsat SR checked against MODIS SR • MODIS SR validated extensively • Landsat 7 SR  MODIS SR (daily) • Same acquisition date, ~30 minutes apart (Landsat 7 images) • Aggregate Landsat to MODIS resolution • Use samples from homogeneous areas • Fully automated, will be applied to all Landsat 7 images • Landsat 5 SR  MODIS SR (NBAR) R square RMSD

  12. SVM Training & Classification Forest change products Landsat Imagery TDA Training Data Automated Forest Change Mapping Method • SVM (support vector machines) more accurate than other classifiers (Huang et al., 2002; Pal & Mather, 2005, and others) • Automation possible by TDA (training data automation) (Huang et al., 2008)

  13. Cloud and Shadow Masking Fig. 1. Use of a cloud boundary in a spectral-temperature space to delineate cloud pixels. In the spectral-temperature space shown in (c), the color of each dot indicates the number of pixels plotted at that point, which increases from purple to blue, green, yellow, and red. The cloud boundary consists three segments, which are labeled from 1 to 3, and are defined and joined by points a, and b. See section 2.4 on discussions of the cloud boundary. (Huang et al., in press)

  14. Cloud and Shadow Masking Landsatimage, R, G, B = bands 5,4,3 cloud mask by the Luo et al. (2008) method Cloud mask by this study (Huang et al., in press) 18/35, 1992-07-21 21/39, 1997-08-25

  15. FCC Products Assessment • Comparison with existing land cover products • Global & regional products • Rule out gross errors • NAFD dense time series change products • Highly reliable • Accuracy estimates • Design-based accuracy assessment

  16. Use Existing LC Products to Rule Out Gross Errors Figure 5. Percent forest cover (PFC) within 10 km grid cells for the 2000 epoch calculated using a TDA-SVM derived FCC product and the MODIS VCF tree cover product. The near 1:1 relationship between the two sets of PCF values is an indicator of the high quality of the TDA-SVM derived FCC product.

  17. Dense Time Series FCC Products State level: Utah, Mississippi, Alabama, Maryland, North Carolina LVIS acquisition sites

  18. Accuracy Assessment Random Sample (known inclusion probability) Before After

  19. Summary • Producing global forest cover change products will be achieved through • Fully automated algorithms • Systematic quality checking/verification • Mass processing capability • Issues with the GLS Data Sets • Major data holes • Difficulty to calibrate the 1990 data set

  20. A Major Data Hole in the GLS 1975 No data for the north part of SA INPE has the data to fill the gap

  21. Eastern Siberia Not Covered in 1990

  22. GLS 1990 Calibration Challenge

  23. Backup Slides

  24. Training Data Automation Forest Index Threshold FI (Huang et al. 2008)

  25. SVM More Accurate • Classify using optimal class boundary • TDA-SVM tested in many places • Huang, C., Song, K., Kim, S., Townshend, J., Davis, P., Masek, J. & Goward, S.N. (2008). Use of a Dark Object Concept and Support Vector Machines to Automate Forest Cover Change Analysis. Remote Sensing of Environment, 112, 970-985.

  26. Coordination with Related Projects • NASA LCLUC • Chandra Giri, USGS EROS • Tropical Mangrove Forests: Global Distributions And Dynamics (1990-2005) • Matthew Hansen, South Dakota State University • Producing Composite Imagery And Forest Cover And Change Characterizations For The Humid Tropics - A Contribution To The MDGLS Activity • Mutlu Ozdogan, University of Wisconsin • Land-Use And Land-Cover Changes In Temperate Forests Of European Russia: The Past, The Current, And The Future • David Skole, Michigan State University • Enhancing Global Observations And Information On Tropical Forest Change Using Landsat Global Data • John Townshend, University of Maryland • Three Decades Of Forest Cover Change In The Americas Evaluated Using The Geocover And MDGLS Data Sets • Xiangming Xiao, University of New Hampshire • Developing Land Cover Classification Products In Monsoon Asia Over The Period Of 2004-2007 Through Integration Of Landsat And ALOS/PALSAR Images

  27. Coordination with Related Projects • Other NASA Projects • Jeff Masek, NASA/GSFC • Recent North American Forest Dynamics via Integration of ASTER, MODIS, and Landsat Reflectance Data • Jeff Masek, NASA/GSFC • Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS)

  28. Stay tuned … • Five-year project started this summer • Current progress • System setup • Data ingestion • Prototype studies

More Related